Importance sampling for the efficient simulation of adaptive systems in frequency nonselective slow Rayleigh fading

Importance

Sampling

for the Efficient Simulation

of Adaptive Systems in

Frequency Nonselective

Slow Rayleigh Fading

w.

A.

Al-Qaq

J.

K. Townsend

~nter

for Communications

and

Signal Processing

Department of Electrical

and

Computer Engineering

---North Carolina State University

TR-94j4

Technical Area: Modeling and Simulation Techniques

Importance Sampling for the Efficient Simulation of Adaptive

Systems in Frequency Nonselective Slow Rayleigh Fading

1

Wael A. AI-Qaq 2 J. Keith Townsend 3

Center for Communications and Signal Processing, Department of Electrical & Computer Engineering, North Carolina State University, Raleigh, NC 27695-7914

Tel: (919)515-7353 Fax: (919)515-5523

Abstract

Importance sampling (IS) is recognized as an efficient technique in reducing the simulation run time needed to estimate low bit error rates (BER's) in digital communication systems. However, IS applications presented in the literature thus far have been primarily limited to systems with additive white Gaussian noise (AWGN).

In this paper, we present an IS stochastic technique for the efficient simulation of adaptive systems which employ diversity in the presence of frequency nonselective slow Rayleigh fading and AWGN. After accounting for the overhead of the optimization algorithm, average speed-up factors of up to 6 orders of magnitude (over conventional Monte Carlo (MC)) were attained for error probabilities as low as 10-11

•

lThiswork was supportedin part by the Center for Communications&Signal Processing, North Carolina State University.

2W. A. Al-Qaq is an IBM Graduate Fellow.

1

Introduction

As the demand for wireless data communications increases, so will the need for low bit error

rate (BER) wireless links. Since the wireless channel is frequently characterized by Rayleigh

fading and AWGN, adaptive reception as well as diversity combining schemes [1, 2, 3] are

techniques useful for mitigating time variations in the channel and of achieving lower BER's

for data applications. Unfortunately, this added complexity makes closed form analysis of

the BER infeasible and renders MC simulation as the primary substitute for performance

evaluation. However, utilizing conventional MC for low BER estimation can in itself be a

prohibitive task due to long run times.

As an alternative,

Me

based IS techniques are frequently applied to significantly reduce

simulation run time for a given estimator precision. This substantial reduction, however,

is normally accompanied by the difficulty of specifying an efficient IS scheme and the

cor-responding optimal parameter settings. For the diversity system considered in here, the

cumulative memory behavior of the adaptive receiver, the decision-directed phenomenon,

and the nonlinearity of the adaptive algorithm render analytical optimization techniques

[4, 5, 6] as ineffective. In addition, it is unclear if and how numerical optimization (large

deviations) techniques [7, 8, 9, 10] are applicable in this case.

The main contribution of this paper is to introduce a stochastic IS methodology for

the efficient simulation of systems characterized by diversity and adaptive receivers in the

presence of nonselective slow Rayleigh fading and AWGN.This IS stochastic gradient descent

(SGD) algorithm, which we first presented in

[11],

is utilized in here to determine the

near-optimal IS parameters that characterize the dominant fading process. The fading model

assumed in this paper obeys a first order Markovian process [1, 12, 13, 14]. Using a simple

illustrative example, we show that for this case, the IS simulation technique is mostly efficient

when the statistics of the fading model obey a modified Markov chain distribution [7,8, 15].

A fourth order diversity system was simulated at three different time instants and using

the corresponding optimal IS parameters. In addition, two signaling formats were considered,

speed-up factors of up to 6 orders of magnitude (over conventional Monte Carlo (MC)) were

attained for error probabilities as low as 10-11 .

2

System Description

Consider a carrier-modulated real-valued signal S(

t)

transmitted over L independent diver-sity channels, where

S(t)

=

Re

{t,

d(k)g(t - kT) exp(jwct)}

(1) where L~od(k)g(t- kT) is the complex lowpass envelope with d(k) being the kth data

symbol (real or complex), and

g(t)

being the impulse response of the transmit filter. We is the carrier frequency in rad/ sec. Each channel introduces a frequency nonselective and slow

Rayleigh fading in addition to AWGN. The equivalent lowpass and complex time-varying

impulse response of each channel is given by [16]

hi (r,t)

==

c;(t )t5(r ), i

==

1, . . . ,L

(2)

where for i

=

1, ... ,L, Ci(t) is a complex-valued Gaussian random process. The slow fading

process is assumed to be constant for the duration of one symbol period [1, 12,16]. Employing

this piecewise constant approximation for the fading process, the received signal over the ith

channel can be expressed as [1, 16]

Xi(t)

=

Re {

l~

ci(k)d(k)g(t - kT)

+

ni(t)] exp(jwct) } ,

i

=

1, ... ,

L

(3)

where [L~oci(k)d(k)g(t - kT)

+

ni(t)] represents the received complex lowpass envelope.

The fading gain of the ith (1

:S

i ::;

L)

fading channel ci(k) is a complex Gaussian random variable (CGRV) with E{Ci(k)} == 0 and E{lci(k)12}

=

20"2. For i

i=

i,

ci(k) and cj(k) are

independent processes. In this paper, the complex fading process ci(k)

==

3.i(k)

+

jbi(k) has

an even Doppler power spectrum [17] which implies that {Ri(k)} and {bi(k)} are independent

Gaussian processes with a zero mean and an identical autocorrelation function R(l), where

The fading model assumed in this paper obeys a first order Markovian process [1, 12, 13, 14],

namely

(5)

The correlation parameter pis a measure of the rate of the fading channel fluctuations. This

parameter is primarily determined by the product of the symbol period T and the fading

bandwidth fd (i.e., the 3dB cutoff frequency of the Doppler power spectrum). In the case of

a first order Butterworth Doppler power spectrum

[18],

the correlation parameter is given

by

(6)

where for a slow Rayleigh fading channel, we have 0 < fdT

«

1. For example, a symbol

rate of 50 KHz and a fading bandwidth of 10 Hz would correspond to p

=

0.9998. Other

expressions for p can be found in [18] for several commonly encountered Doppler power

spectra.

The autocorrelation function in (5) yields the following state equation

(7)

where for i

=

1, . . .

,L,{Wi(k)}

is a sequence ofiid zero-mean CGRV's with E{lwi(kW}

=

2er

2(1

_-

_{p2), and for i} =1= j,

{wi(k)}

and

{wj(k)}

are independent processes.

The complex additive noise at the ith antenna is given by

ni(t),

where each

ni(t)

is a complex AWGN process, with

E{ni(t)}

=

0 and

E{ni(t)niH(t

+

T)}

=

No

5(T) (H

denotes the complex conjugate), with

ni(t)

and

nj(t)

being independent processes for i =1= j.

The received signal is demodulated by a local carrier exp(

-jwct),

match-filtered by

g(T-t), and sampled every T seconds. Thus, the resulting ith output (input to the ithequalizer)

during the kth symbol period is given by

~(k)

=

ci(k)d(k)

+

ni(k), i == 1, ... ,L

(8)

In the above equation, it is assumed that the transmit filter is normalized (i.e.,

J{

Ig(t)\2dt

=

1), therefore, E{l_ni(k)12 } =

No-

A block diagram of the the digital communication system

exp (-j We t )

exp(-j ~ct)

Update Algorithm

Update Algoritlun

"

d

Figure 1: A block diagram of a digital communication system with diversity and adaptive filtering in the presence of nonselective slow Rayleigh fading and AWGN.

A partially coherent reception technique was employed in

[1, 12]

using a decision-directed

adaptive Kalman filter. This receiver assumes full knowledge of the statistics of the incoming

signal. In this paper, a decision-directed recursive least square [19, 20] adaptive receiver is utilized to compensate for the fading process on the ith diversity. this partially coherent

algorithm does not assume any prior knowledge of the statistics of the input signal, and is

therefore more practical to implement. Define the random data vector at instant k as

D(k)

=

[d(k),d(k-1), ... ,d(O)],

the ith AWGN vector at the output of the receive matched-filter at instant k as

and the corresponding vector

N(k)

of the iid random vectors

{Ni(k)}t:l

as

where

and

and let C(k) be the corresponding vector of the iid random vectors {C_{i (}k )}~1' namely

C(k)

=

A(k)

+

B(k)

==

[C1(k),C2(k),... , CL(k)] (9)

where A(k) and B(k) are defined in a fashion similar to C(k). In addition, define the received

random vector at instant k as

and the corresponding vector X(k) of the iid random vectors {X.,: (k)}f:l as

(10)

The output of the ith adaptive receiver for the kth transmitted symbol d(k) (k

2:

k_{o )} is

given by [19, 21]

y.,:(k)

i=l, ... ,L(ll)

where ko is the number of symbols used to train the equalization algorithm (training

se-quence), and 5 is a small positive number. 0

<

A ::;1represents the forgetting factor of the

algorithm. The purpose of this factor is to weight the most recent symbols more heavily and

thus allow the equalizer to track time variations in the channel. The input to the decision

device is given by

L

y(k) LYi(k)

i=l

L

Lg(~(k))

=

G(X(k))

i=l

At the end of the training period, a decision-directed adaptive algorithm is employed whereby

decisions made on the output sequence {y(

k)}

(i.e., {d(

k)})

are used to replace the actual transmitted symbol sequence {d(k)}. This will effectively aid the adaptive receiver in

track-ing channel variations, but may increastrack-ingly hinder the tracktrack-ing capability with time due

error propagation. This error propagation effect will be demonstrated later on when

consid-ering a fourth order diversity example.

3

IS

Formulation

The adaptive system considered in the previous section clearly possesses a cumulative

mem-ory behavior as evident from

(11).

This behavior is quite common to all adaptive algorithms where the decision made at each time instant k utilizes a weighted version of all the past received samples up to time zero. In addition to this behavior, the decision-directed

phe-nomenon, the strong correlation present in the input signal, and the non-linearity of the

adaptive algorithm make closed form analysis of the BER infeasible. As an alternative,

Me

simulation is commonly applied to estimate the BER performance.

Since the statistical distribution of the complex random vector X(k) is dependent on the

correlation parameter p and the power in the fading process 20"2, it would be appropriate to

denote its corresponding probability density function (pdf) as fX(k)(JY(k), 0), where

and

6-E>

=

[p , 0-]

fC(Ie) (C(k),8 )fN(k)(N( k))fD(k)(D( k))

fA(lc)(A( k),0 )fB(lc)(B(k),0 )fN(lc)(N(k) )fD(lc)(D(k))

(13)

(14)

For the fading model assumed in (5), the distribution of the ith independent random vectors

Ai(k) and Bi(k) are identical with the ith marginal pdf fA;(lc)(Ai(k), 0) being given by

Ie-I

(15)

1 {

a~(O)

}

(27l"u2)(k+1)/2(1 _ p2 )k/2exp - 2u2 x

ex {_

L:j:ci(ai(j

+

1) -

pai(j))2}

p 2(1 _ p2)U2 , i

=

1,. · ·,L

Observe that the notation fX(k)(JY(k),8) was only chosen to explicitly show the dependence

on (3 since for most practical cases and cases where IS is applicable (i.e., high signal to

noise ratio

(SNR)),

the slow Rayleigh fading process is the dominant factor. Therefore, we

consider biasing only the fading processes A(k)and B(k)leaving the AWGNrandom process

N( k)

unbiased. In addition, the parameter vector

e

is related to the biasing scheme which

we adopt and address in the next section.

Let the expectation with respect to the marginal pdf fX(k)(~y(k), e) be denoted as

Ek,e{.}.

The subscript k is used to emphasize the time-dependent nature of the

expec-tation. In addition, let I(G(X( k)) be equal to 1 when the detected symbol is in error and

zero otherwise, where G(X(k)) is given by

(12).

The probability of making a decision error

at some time instant kis given by the following instantaneous BER expression

(16)

where 10(k) is the indicator function of the set Q(k)

=

{X(k) : I(G(X(k)))

=

1}.

To apply IS, note that

(16)

can be written as

(17)

whereE_Ic,eO(k){ .} denotes the expectation operator with respect to the IS pdffX(k)(X( k), e*( k))

with the time-dependent parameter vector

and

WX(k)

(X(

k),

e,

e

* (k))

e

* (k)

=

[p

* (k) ,0"* (k)]

=

WA(k)(A(k),

e,

e*(k))WB(k)(B(k),

e,

e*(k)) fA(Ic)(A(k),

e)

fB(Ic)(B(k),

e)

- fA(k)(A(k), e*(k)) fB(k)(B(k), e*(k))

(18)

Observe that the random vectors A(k) and B(k) are iid with

L

fA(k)(A(k),8) ==

II

fAi(k)(Ai(k),0)

i=l

and

L

fA(k)(A(k), 8*(k)) ==

II

fAi(k)(Ai(k), E)*(k))

i=l

Upon incorporating the preceding two equations into (19), we get

(20)

(21)

L

WX(k)(X(k),

e,

e*(k))

=

II

WA.(k)(Ai(k),

e,

e*(k))WB;(k)(Bi(k),

e,

8*(k)) (22)

i=l

where WAi ( k ) ( Ai(k),8,8*(k)) and WB i(k)(Bi(k),E),8*(k)) have an identical distribution,

with

w . (A·(k) 8 e*(k))

=

fA.(k)(Ai(k), e)

A.(k) , , , _{f A.(k)( Ai( k), 8*( k))}

A Me-based IS estimator of (17) is given by

A 1 Nx

p(k,e)

=

N

:L

I(G(X (k,i )))wX(k)(X (k,i ), 8 , 8 *(k))

X i=l

(23)

(24)

It is straight forward to show that the estimator in

(24)

is an unbiased estimator of P( k,8)

(i.e., Ek.e.Ck){F(k,

en

=

P(k,8)). The variance of the IS estimator in

(24)

is given

V{F(k,8)} == V(k,8,8*(k))

- ;x

Vk,eOCk) {lnCk)wXCk)(X(k),8,e*(k))}

(25)

where V : R2 _{X ~2 ~ [0,00), and VJc,e.(k){.} denotes the time-dependent variance}

opera-tor with respect to the simulation pdf fX(Jc)(X(k),0*(k)). The empirical precision of the

estimator in (24) may be found by using the sample variance estimator

A A 1 Nx . * P2(k,8)

V{P(k,8)}

=

- 2

:L

I(G(X (k,i )))w5CCk)(X (k,1. ), 8 , e (k)) - N (26)

Nx i=l x

For a given relative precision ao > 0, the simulation is terminated when the condition

3.1

The IS Biasing

Scherne

Two important issues are worth addressing at this point. First, note that for i = 1, ... ,

L,

the in-phase and quadrature Gaussian processes Ai (k) and Bi (k) are identically modified

according to the parameter vector e*(

k).

This is the case since A_{i (}

k)

and B_i(

k)

are equally

weighted in the real and imaginary parts ofYi(k) in (11). In addition, each of the random

processes {Ai

(k )}t:l

_and

{B

i

(k )}f: l

must be equally biased as well since according to (12),

the output

y(

k) is the sum of the equally weighted iid outputs {Yi( k)}f=l. This implies that

the modification of the underlying density function fX(k)(X(k), e) is entirely governed by

the parameter vector 8*(k ).

Second, observe that altering the parameter vector 0 implies that the adopted IS

sim-ulation technique is performed under a modified Markov chain distribution [15, 8, 7]. At a

given time instant k, this technique effectively transforms the original unbiased

autocorrela-tion funcautocorrela-tion R(l) of (5) into

R*(l,k)

==

(T*2(k)p*lll(k), 0

<

p*(k)

<

1

₍₂₇₎

In the simple case of a memoryless nonlinearity followed by an integrate and dump receiver, and when the input samples have a Markov chain distribution, it was shown [7]

that the above simulation technique is more efficient than translating the mean

[4, 10]

or

scaling the variance [5,6] of the iid random variables {wi(k)} in (7). The ineffectiveness of

the of the mean translation and variance scaling techniques is primarily attributed to the

fact that the random process X(k) has a zero mean and X(k)

==

0 E f2(k).

This can be illustrated by Considering the system discussed in the previous section with

a first order diversity and binary signaling (d(k)

==

±1). Furthermore, consider the time

instant k

=

1, and let the SNR

==

00 (i.e., N1(1)

==

0) along with the quadrature random

a1(l)

Figure 2: A contour plot of the original pdf,

fA

_{1(1)( A 1}(1), 0 ) and the corresponding error region.

A plot of the contours of

fA

1(1)( A 1

(1),

0) and the error region (E1

U

E2 ) for 0

=

[0.99 , 1]

is shown in Fig. 2. The probability of error in this case is equal to P(1,0) = 4.51 X 10-2 •

Note that as the value of p ---+- 1, the contours area over the region E₁

U

E₂ will gradually

vanish and as a result, P(l, 0) ---+

o.

This should be the case since it can be shown that

1

[~1

-

p2]

P(1,0)=;tan-1 p

Clearly, limp--+l P(l,0)

=

o.

Even though translating the contours diagonally into E₁ or E₂ will increase the raw (i.e.,

unweighted) error count, the variance of the IS weight function over the entire error region

(E1

U

E2 ) will rather increase since the modified pdf can only concentrate its mass on either

El or E2 but not both. Uniform variance scaling (i.e., biasing (72 only) will, on the other

hand, proportionally scale the contours in all dimensions and the net increase in the raw error count will be zero. This conclusion can also be deduced from the fact that P(l,0), as

given above, is independent of a,

The difficulty encountered with the above two schemes can be surmounted by modifying

a1(1)

Figure 3: A contour plot of the modified pdf,

fA

₁(1)( A 1(1),

e~t(1)), e~t(1)

[0.4674 , 0.0211].

realize a good IS biasing scheme, namely increasing the raw error count and evenly distribute

the weight function over the error region E₁

U

E₂•

The IS variance (Eq. (25)) for this simple case can be attained in a closed form and

minimized to yield e~t(l). Instead, the search for e~t(1) was conducted stochastically

using the SGD algorithm, which we first presented in [11], with

e

being the starting point.

The search yielded 8*(1) ~ 8~t(1)

=

[0.4674 , 0.0211], and a plot of the contours of the

modified pdf is shown in Fig. 3.

4

Optimizing the IS Distribution

In

order to attain a significant reduction in the number of detector decisions Nx needed

to estimate P(k,0) with a given accuracy, the optimal parameter setting 8~t(k) which

minimizes the variance in (25) must be determined. Since a closed form expression of (25) is

not available, analytical optimization techniques [4, 5, 6] are rendered ineffective. In addition,

as we illustrated earlier, having a zero mean random process

X(k)

and

X(k)

=

0 E

n(k)

adaptive algorithm make it unclear how numerical optimization (large deviations) techniques

[10, 7, 8, 9, 10] can be modified to accommodate the current application.

As an alternative, the IS SGDalgorithm which we first presented in [11, 22] is utilized to

search for the optimal IS parameter vector 8~t(k)in a two dimensional space. Insummary,

the SGD algorithm is based on the fact that for any vector 8~t(k) to be a local or global

minimum of V(k, 8,0*(k)) in

(25),

the following equation must hold

(28)

Thus, the search for the optimal IS setting is conducted by the SGD algorithm in a descending

fashion according to

0*(k,

n

+

1)

= 8*(k,

n) - (3(n)\7

e·(k)V(k,0,8*(k))le·(k)=e·(k,n) (29)

where

(3(

n)is the step size taken at the nth iteration, and V" e·(k)V( k,8,8*(k)) is an unbiased

estimate of the left-hand side of

(28).

It was shown

[11],

that such an estimate can be

constructed using the following estimator

VeO(k)V(

e,

e*(k))

=

~

I:

I(

G(X(k, i)))WX(k)(X(i, k),

e,

e*(k)) X

Nx ;=1

\7eO(k)WX(k)(X(k, i),

e,

e*(k)) (30)

In order to implement the above estimate, we need to compute the gradient of the IS weight

function.

* [8WX(k)(X(k),

e,

e*(k)) 8WX(k)(X(k),

e,

e*(k))] (31)

\7eO(k)WX(k)(X(k),

e,e

(k))

=

8p*(k) , 8(j*(k)

Introducing the following definitions will help simplify the expression of the above

gra-dient. Let for i

=

1, .. ·

,L,

6. p*(k)(j*2(k)I:j~ci(ai(j

+

1) -

p*(k)ai(j))2

R1(Ai(k), e*(k)) = (1 _ p*2(k))2(j*4(k)

(1 - p*2(k))(j*2(k)E;':~(ai(j

+

l)ai(j) - p*(k)ai(j)2)

R_2(Ai(k),8*(k))

~ L:j::~(ai(j

+

1) - p*(k)ai(j))2 (33)

(1 - p*2(k))u*3(k))

with R1(Bi(k),e*(k)) and R2(Bi(k),e*(k)) being defined in a fashion similar to

(32)

and (33)

respectively. After taking the derivative of both sides in

(22)

and utilizing the above

defini-tions, it can be shown that

8WX(k)(X(k),8,8*(k))

8p*(k)

and

8WX(k)(X(k),0,e*(k))

80-*(k)

{

2L(k + l ) }

•

=

0"* (k) WX( k) (X (k),8, 8 (k))

-L [(

al

(0)

+

b~

(0))]

* ~ 0"*3 (k) WX( k) (X (k),8, 8

(k))-L

L

[R2(Ai(k), 0·(k))

+

R2(Bi(k), 8*(k))] x

i=l

WX(k)(X(k),

e,

e*(k))

(35)

Observe that during the nth iteration of the SGD algorithm in

(29),

the IS estimators

of all the three quantities P(k,8), V{P(k,8)}, and "Ve.(k)V{ P (k, 8 )} are obtained by

sampling from the simulation pdf at the nth iteration, namely fX(k)(X(k),8*(k,n)). In

addition to yielding an optimal estimator of P(k,8) asn - 00, this approach will generally

provide suboptimal estimates of V{P(k,

e)}

and "Ve- V{p}, which are sufficiently accurate

to successfully perform the SGn algorithm.

Let the SNR per diversity be defined as

E{ld(k)Ci(k)12

} E{ld(k)12} 20-2

I = E{ni(k)} = No i = 1, ...,L.

(36)

For a given diversity order 1 ::;i ::; L ,a SNR,and a time instant k

2:

ko , the SGn algorithm

is applied for some value of 8 such that P(k,8) rv 10-2• this technique will help circumvent

(Nx ~ 1000) can be sufficient to accurately compute 'V'e.(k)V{p(k,8n!e.(k)=e.(k,l). The

optimal parameter vector e~t determined for a high P(k,8) is then used to choose a starting

point 8*(1) at higher values of P(k,8). This "extrapolation" technique was shown to be

very effective in a variety of practical digital communication and queing systems applications

[11,22] and for highly nonlinear systems with large dimensionality

[23].

with this technique

applied efficiently, the overhead (in number of decisions) involved in determining 8~t(k) will

be insignificant compared to the savings in number of decisions Nx needed to accurately

estimate a low P(k,8).

As an example, consider a second, and a fourth order diversity systems with BPSK

signaling and a per-diversity SNR of I

==

30 dB. The equalizer was trained using a fixed

training sequence of 12 symbols and the BER was measured 10 symbols after training (i.e.,

P(22,8)). Due to the absence of intersymbol interference (lSI) and the symmetry of the

signal constellation about the origin, the probability of error is insensitive to the choice of

the data sequence after the training period. This effectively results in an optimal IS setting

that is sequence-independent, as was empirically verified. In each case, 8 was chosen to

attain a high P(22,8). Since the BER being considered is high, the improvement will not

be significant, and the optimal vector is identified by the convergence of the SGD algorithm.

The search for the near-optimal vector 8~t(22) was executed iteratively using the SGD

algorithm with

Nx

=

1000 decisions per estimate of 'V'e.(k)V{P(k,

8n

per iteration. The step size

f3(

n) at the nth iteration was chosen according to

~

(3(n)

== , .

-IIV

{P( k,e)}!ee(k)=ee(k,n)

II

(37)

On the nth iteration, this step size will result in a maximum incremental or decremental

change of .6. to the components of the parameter vector 8*(k,n). As observed in

[11,23],

selecting a small .6. will result in a slow convergence but a higher accuracy in locating the

optimal setting as opposed to a larger .6. that yields a faster convergence rate but may cause

some deviation from the the correct path of the gradient descent. Typical values of ~ may

Diversity (L) 0 _0~t(22) P(22,0) Raw count

1 [0.992, 0.7071] [0.9637 , 0.3295] 4.34 X 10-2 17.3%

2 [0.98, 0.7071] [0.9602, 0.504] 2.78 X 10-2 8.75%

4 [0.96, 0.7071] [0.947, 0.6042] 1.19 X 10-2 _2.7%

Table 1: The optimal IS parameters for k == 22. The signaling format is BPSK andI == 30dB.

0.5 :.~

/ \\R(I)

//~4

\\00\.

-50 -30 -10 10 30 50 0)

Figure 4: A plot of the original and optimal IS autocorrelation functions for BPSK. L

=

4,

e

= [0.96 , 0.7071], and I = 30 dB.

to p, and C= 4 X 10-5 • The results are shown in Table 1. e*(k) :::::: e~t(k,n). For each

case considered in Table 1, the optimal IS autocorrelation function R~t(l, k) yields a less

correlated fading signal, as compared to R(l), with a reduced mean power (i.e., less SNR).

This effectively translates into an increase in the Doppler frequency bandwidth (i.e., faster

fading) and a reduction in the energy of the fading power spectrum. A plot of the original

and modified autocorrelation functions for the case L

=

4 is shown in Fig. 4.

Two important observations can be deduced from Table 1. First, the raw error count at

the optimal IS setting decreases as the diversity order increases. Moreover, it was

experi-mentally observed that the error count corresponding to the different diversities in Table 1

is roughly maintained at the same level for different values of p and k.

equation

(38)

where k

==

22. In fact, as our empirical results will show later, the above equation holds

for other time instants as well, and its accuracy increases as the correlation parameter p

approaches unity. Clearly, the above equation can be exploited to transform a 2-D search

into a 1-D search, however, we only chose to utilize the above equation in identifying a good

starting point prior to conducting the SGD search.

Thus, for a given SNR, a correlation function R(1)

==

(1"2pili, a time instant k1

2:

ko ,

and a diversity order L, the near-optimal parameter setting 8~t(k), or equivalently, the

near-optimal autocorrelation function R~t(l,k1 ) is determined as follows:

• Choose a p'

<

p and a _<7'

==

(J" such that P(k1 ,8') ~ 10-2•

• U sing the IS SGD algorithm, determine the near-optimal setting for the system with

8'

==

[p' ,

<7'],

• Choose a starting point 8*(k1, 1) such that

(39)

where

(40)

• Perform a 2-dimensional search using the SGD algorithm to locate e~t(kl)

The optimal setting e~t(kl) can then be used as the starting point of the SGD algorithm

to locate the optimal parameters at a time instant k2

>

k1 , and so on.

4.1

A Fourth Order Diversity Example

Consider a fourth order diversity system

(L

= 4). The simulation algorithm discussed in the previous section was applied to simulate the HER at the time instants, k1

=

_22, k2

=

_32,

and k

k _0~t(k) P(k,0) V{F(k,0)}

Sp(k)

Raw Count 22 [0.9708, 0.0417] 1.894X 10 11 1.852 X 10-2 2 _{8.236 X 10}7 _2.28%

32 [0.9753, 0.0455] 4.567X 10 11 2.372 X 10-21 1.53 X 107 2.9%

42 [0.9802, 0.0503] 1.621 X 10 10 2.135 X 10-19 _{1.6X 10}5

2.82%

Table 2: The optimalIS parameters and the corresponding estimated probabilities, variances, and speed-up factors. 0

=

[0.9999, 0.7071] andI

=

30 dB, and the signaling format is BPSK.

k _0~t(k) P(k,0) V{P(k,0)}

Sp(k)

Raw Count

22 [0.9873, 0.0631] 2.74 X 10-9 4.97 X 10-17 1.82 X 104 2.96%

32 [0.9910, 0.0745] 5.64 X 10-9 3.46 X 10-1 7 _{1.169 X 10}5 _2.7% 42 [0.9921 , 0.0792] 1.725 X 10-8 1.56 X10-15 4.1X 103 3.11%

Table 3: The optimalIS parameters and the corresponding estimated probabilities, variances, and speed-up factors. 0 = [0.9999, 0.7071], I = 30 dB, and the signaling format is QPSK.

was always normalized to unity (E{lci(k

)1

2

} = 1). The symbol rate was set to 50 KHz, the fading bandwidth fD was assumed to be 2.2 Hz which corresponds to p

=

.9999, and a SNR

of I

=

30 dB was considered. In addition, two signaling formats were considered, BPSK and

QPSK. In each case, the forgetting factor of the adaptive algorithm A was set equal to the

correlation parameter p and 5 = 4 X 10-5• The starting point for the first entry in Table 2 was 0*(22,1) = [.9476, .0313] which was determined using Table 1 and Eq. (40). In addition,

for the given 0, " and k, the near-optimal IS parameters determined for the BPSK case

were also used as a starting point for the QPSK case. In each case, the step size was chosen

according to (37) and the values of ~ ranged from 0.0003 to 0.001. Another interesting case

to consider is when /

=

00 since it represents the irreducible BER performance. This case

was simulated for QPSK and 0

=

[0.9999 , .7071], and the results are shown in Table 4.

The estimates of the variance V{"p(k,0n and the instantaneous BER P(k,0) were

computed using an ensemble of NE

=

50 estimates of Nx

=

1000 decisions per estimate.

The time-dependent speed-up factor Sp(k), corresponding to V{..P(k, 0n and the estimator

in (24), was calculated according to

k _0~t(k) P(k,0)

V

{.P(k, 0)} Sp(k) Raw Count

22 [0.9879 , 0.0635] 1.89 X 10 9 2.15 X 10-17 _{3.17 X 10}4 _1.17%

32 [0.9910, 0.0745] 3.82 X 10 9 2.31 X 10-17 1.15 X 10-5 1.56%

42 [0.9922, 0.0804] 1.3368 X 10 8 8.205X 10-16 6.134 X 103 2.09%

Table 4: The optimalISparameters and the corresponding estimated probabilities, variances, and

speed-up factors. 0 = [0.9999, 0.7071], /

=

00, and the signaling format is QPSK.

where NM C is the conventionallVIC number of decisions required to attain the same accuracy

as our IS scheme. NM C was computed based on a 95% confidence interval [24]. The overhead

(in number of decisions) was not included in the computations of the speed-up factor. The

reduction in speed-up factors due to overhead ranged from 1 to 2.5 orders of magnitude.

This reduction is clearly dependent on the choice of ~.

Several interesting observations can be made from the results In Tables 2, 3, and 4. First note the increase in P(k,0) as time (k) increases. This increase is due to the error

propagation effect of the decision-directed algorithm ( i.e., feeding back erroneous decisions

while updating the adaptive algorithm). As time increases, P( k,e) will also increase until it eventually reaches an intolerable rate, and transmitting a new training sequence would

become necessary to improve the adaptation algorithm. This signaling technique which

periodically interleaves training and data sequences is frequently used by adaptive algorithms

[2, 25]. Inaddition, observe that the increase in P( k,

e)

is also accompanied by a decrease in

Sp( k) as would have been expected. Any further decrease in Sp( k) might also be attributed

to an increase in the memory of the algorithm as k increases.

Another two important observations pertaining to the optimal IS setting e~t(k)are the

decrease in the bias as k increases, and how well

e:,

i

k ) satisfies Eq. (38). A plot of the optimal trajectory (p~t(k), O";"t(k)) in Table 2 vs. time is shown in Fig. 5.

5

Conclusion

In this paper, we presented a stochastic IS methodology for the efficient simulation of

n'" 'k) t""opt(

0.982

0.98 k:= 42

0.978

0.976

k=32

0.974

0.972

k =22

0.05 0.052 0.042 0.044 0.046 0.048

0.97 '---~ 0.04

Using a first order Markovian model for the fading channel, the search for the optimaly

modified Markov distribution was accomplished by the robust SGD algorithm. Extending

this technique to accommodate adaptive algorithms other than the RLS algorithm is straight

forward. Speed-up factors of up to 6 orders of magnitude over conventional MC were achieved

for error probabilities as low as 10-11 _{after accounting for overhead of the SGD algorithm.}

References

[1]

Pooi Yuen Kam. Adaptive Diversity Reception Over a Slow Nonseleetive Fading

Chan-nel. IEEE Trans. Commun., 35(5):572-574, May 1987.

[2] N. W. K. Lo, D. D. Falconer, and A. U. H. Sheikh. Adaptive Equalization and Diversity Combining for Mobile Radio Using Interpolated Channel Estimates. IEEE Trans. Veh.

Technol., 40(3):636-645, Aug. 1991.

[3] P. Balaban and J. Salz. Dual diversity Combining and Equalization in Digital Cellular

Mobile Radio. IEEE Trans. Veh. Technol., 40(2):342-354, May 1991.

[4] D. Lu and K. Yao. Improved Importance Sampling Technique for Efficient Simulation of Digital Communication Systems. IEEE J. Select. Areas Commun., 6(1), Jan. 1988.

[5] K.S. Shanmugan and P. Balaban. AModified Monte-Carlo Simulation Technique for the

Evaluation of Error Rate in Digital Communication Systems. IEEE Trans. Commun.,

COM-28(11):1916-1924, Nov. 1980.

[6]

B. R.

Davis. An Improved Importance Sampling Method for Digital Communication

System Simulations. IEEE Trans. Commun., COM-34(7):715-719, Jul. 1986.

[7] R. K.

Bahr and J. A. Bucklew. Quick Simulation of Detector Error Probabilities in the

Presence of Memory and Nonlinearity. IEEE Trans. on Commun., 41:1610-1617, Nov.

[8] J. A. Bucklew, P. Ney, and J. S. Sadowsky. Monte Carlo Simulation and Large

Devia-tions Theory for Uniformly Recurrent Markov Chains, J. App.I Pro.,b 27 44 59 M: - , are

1990.

[9]

J.

S. Sadowsky and J.

A.

Bucklew. On Large Deviation Theory and Asymptotically

Efficient Monte Carlo Estimation. IEEE Trans. Inform. Theory, IT-36(3):579-588, May

1990.

[10] J-C. Chen, D. Lu, J. S. Sadowsky, and K. Yao. On Importance Sampling in

Digi-tal Communications - Part I: Fundamentals. IEEE J. Select. Areas in Commun.,

11(3):289-299, Apr. 1993.

[11] W. AI-Qaq, M. Devetsikiotis, and J. K. Townsend. Simulation of Digital Communication

Systems Using a Stochastically Optimized Importance Sampling Technique. In Proc.

IEEE Global Telecom. Conf., GLOBECOM '93, Houston, Dec. 1993.

[12] P. Y. Kam. Optimal Detection of Digital Data Over the Nonselective Rayleigh Fading

Channel with Diversity Reception. IEEE Trans. Commun., 39(2):214-219, Feb. 1991.

[13] Pooi Yuen Kam. Optimal Detection of Digital Data Over the Nonselective Rayleigh

Fading Channel with Diversity Reception. IEEE Trans. Commun., 39(2):214-219, Feb.

1991.

[14] M. Stojanovic,

J.

G. Proakis, and

J.

Catipovic. Analysis of the Performance of a

Decision Feedback Equalizer on Fading Multipath Channels in the Presence of Channel

Estimation Errors. In Proc. ofCISS, Chicago, June 1993.

[15] J. A. Bucklew. Large Deviation Techniques in Decision, Simulation, and Estimation.

New York: John Wiley & Sons, 1990.

[16] John

G.

Proakis. Digital Communications. New York: McGraw-Hill, 1989.

[18]

Pooi Yuen Kam. Bit Error Probabilities of MDPSK Over the Nonselective Rayleigh

Fading Channel with Diversity Reception. IEEE Trans. Commun., 39(2):220-224, Feb.

1991.

[19] S. S. Haykin. Adaptive Filter Theory. Englewood Cliffs, New Jersey: Prentice-HaIl,

1986.

[20] E. Eleftheriou and D. D. Falconer. Tracking Properties and Steady-State Performance

of RLS Adaptive Filter Algorithms. IEEE Trans. Acoust., Speech, Signal Processing,

34(5):1097-1109, Oct. 1986.

[21]

John G. Proakis and Dimitris G. Manolakis. Digital Signal Processing. New York:

Macmillan, 1988.

[22] M. Devetsikiotis, W. Al-Qaq, J. A. Freebersyser, and J. K. Townsend. Stochastic

Gra-dient Techniques for the Efficient Simulation of High-Speed Networks Using Importance

Sampling. In Proc, IEEE Global Telecom. Con!., GLOBECO!v1 '93, Houston, Dec. 1993.

[23] W. Al-Qaq and J. K. Townsend. AStochastic Importance Sampling Methodology for the Efficient Simulation of Digital Communication Systems with RLS Adaptive Equalizers.

To be presented at IEEE Int. Con]. Commun., ICC

'94.

[24] M. C. Jeruchim. Techniques for Estimating the Bit Error Rate in the Simulation of

Digital Communication Systems. IEEE J. Select. Areas Commun., SAC-2(1):153-170,

Jan. 1984.

[25] Pooi Yuen Kam and Cho Huak Teh. Reception of PSK Signals Over Fading Channels

Via Quadrature Amplitude Estimation. IEEE Trans. Commun., 31(8):1024-1027, Aug.